AkitoP commited on
Commit
22abbaf
1 Parent(s): adf0887

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +931 -0
app.py CHANGED
@@ -1,3 +1,934 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  import os
2
  import sys
3
  # to avoid the modified user.pth file
 
1
+ import os,shutil,sys,pdb,re
2
+ version="v2"#if sys.argv[-1]=="v2" else"v1"
3
+ os.environ["version"]=version
4
+ now_dir = os.getcwd()
5
+ sys.path.insert(0, now_dir)
6
+ import json,yaml,warnings,torch
7
+ import platform
8
+ import psutil
9
+ import signal
10
+
11
+ warnings.filterwarnings("ignore")
12
+ torch.manual_seed(233333)
13
+ tmp = os.path.join(now_dir, "TEMP")
14
+ os.makedirs(tmp, exist_ok=True)
15
+ os.environ["TEMP"] = tmp
16
+ if(os.path.exists(tmp)):
17
+ for name in os.listdir(tmp):
18
+ if(name=="jieba.cache"):continue
19
+ path="%s/%s"%(tmp,name)
20
+ delete=os.remove if os.path.isfile(path) else shutil.rmtree
21
+ try:
22
+ delete(path)
23
+ except Exception as e:
24
+ print(str(e))
25
+ pass
26
+ import site
27
+ site_packages_roots = []
28
+ for path in site.getsitepackages():
29
+ if "packages" in path:
30
+ site_packages_roots.append(path)
31
+ if(site_packages_roots==[]):site_packages_roots=["%s/runtime/Lib/site-packages" % now_dir]
32
+ #os.environ["OPENBLAS_NUM_THREADS"] = "4"
33
+ os.environ["no_proxy"] = "localhost, 127.0.0.1, ::1"
34
+ os.environ["all_proxy"] = ""
35
+ for site_packages_root in site_packages_roots:
36
+ if os.path.exists(site_packages_root):
37
+ try:
38
+ with open("%s/users.pth" % (site_packages_root), "w") as f:
39
+ f.write(
40
+ "%s\n%s/tools\n%s/tools/damo_asr\n%s/GPT_SoVITS\n%s/tools/uvr5"
41
+ % (now_dir, now_dir, now_dir, now_dir, now_dir)
42
+ )
43
+ break
44
+ except PermissionError:
45
+ pass
46
+ from tools import my_utils
47
+ import traceback
48
+ import shutil
49
+ import pdb
50
+ import gradio as gr
51
+ from subprocess import Popen
52
+ import signal
53
+ from config import python_exec,infer_device,is_half,exp_root,webui_port_main,webui_port_infer_tts,webui_port_uvr5,webui_port_subfix,is_share
54
+ from tools.i18n.i18n import I18nAuto
55
+ i18n = I18nAuto()
56
+ from scipy.io import wavfile
57
+ from tools.my_utils import load_audio
58
+ from multiprocessing import cpu_count
59
+
60
+ # os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1' # 当遇到mps不支持的步骤时使用cpu
61
+
62
+ n_cpu=cpu_count()
63
+
64
+ ngpu = torch.cuda.device_count()
65
+ gpu_infos = []
66
+ mem = []
67
+ if_gpu_ok = False
68
+
69
+ # 判断是否有能用来训练和加速推理的N卡
70
+ ok_gpu_keywords={"10","16","20","30","40","A2","A3","A4","P4","A50","500","A60","70","80","90","M4","T4","TITAN","L4","4060","H"}
71
+ set_gpu_numbers=set()
72
+ if torch.cuda.is_available() or ngpu != 0:
73
+ for i in range(ngpu):
74
+ gpu_name = torch.cuda.get_device_name(i)
75
+ if any(value in gpu_name.upper()for value in ok_gpu_keywords):
76
+ # A10#A100#V100#A40#P40#M40#K80#A4500
77
+ if_gpu_ok = True # 至少有一张能用的N卡
78
+ gpu_infos.append("%s\t%s" % (i, gpu_name))
79
+ set_gpu_numbers.add(i)
80
+ mem.append(int(torch.cuda.get_device_properties(i).total_memory/ 1024/ 1024/ 1024+ 0.4))
81
+ # # 判断是否支持mps加速
82
+ # if torch.backends.mps.is_available():
83
+ # if_gpu_ok = True
84
+ # gpu_infos.append("%s\t%s" % ("0", "Apple GPU"))
85
+ # mem.append(psutil.virtual_memory().total/ 1024 / 1024 / 1024) # 实测使用系统内存作为显存不会爆显存
86
+
87
+ if if_gpu_ok and len(gpu_infos) > 0:
88
+ gpu_info = "\n".join(gpu_infos)
89
+ default_batch_size = min(mem) // 2
90
+ else:
91
+ gpu_info = ("%s\t%s" % ("0", "CPU"))
92
+ gpu_infos.append("%s\t%s" % ("0", "CPU"))
93
+ set_gpu_numbers.add(0)
94
+ default_batch_size = int(psutil.virtual_memory().total/ 1024 / 1024 / 1024 / 2)
95
+ gpus = "-".join([i[0] for i in gpu_infos])
96
+ default_gpu_numbers=str(sorted(list(set_gpu_numbers))[0])
97
+ def fix_gpu_number(input):#将越界的number强制改到界内
98
+ try:
99
+ if(int(input)not in set_gpu_numbers):return default_gpu_numbers
100
+ except:return input
101
+ return input
102
+ def fix_gpu_numbers(inputs):
103
+ output=[]
104
+ try:
105
+ for input in inputs.split(","):output.append(str(fix_gpu_number(input)))
106
+ return ",".join(output)
107
+ except:
108
+ return inputs
109
+ pretrained_sovits_name="GPT_SoVITS/pretrained_models/s2G2333k.pth"
110
+ pretrained_gpt_name="GPT_SoVITS/pretrained_models/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt"
111
+ def get_weights_names():
112
+ SoVITS_names = [pretrained_sovits_name]
113
+ for name in os.listdir(SoVITS_weight_root):
114
+ if name.endswith(".pth"):SoVITS_names.append(name)
115
+ GPT_names = [pretrained_gpt_name]
116
+ for name in os.listdir(GPT_weight_root):
117
+ if name.endswith(".ckpt"): GPT_names.append(name)
118
+ return SoVITS_names,GPT_names
119
+ SoVITS_weight_root="SoVITS_weights"
120
+ GPT_weight_root="GPT_weights"
121
+ os.makedirs(SoVITS_weight_root,exist_ok=True)
122
+ os.makedirs(GPT_weight_root,exist_ok=True)
123
+ SoVITS_names,GPT_names = get_weights_names()
124
+
125
+ def custom_sort_key(s):
126
+ # 使用正则表达式提取字符串中的数字部分和非数字部分
127
+ parts = re.split('(\d+)', s)
128
+ # 将数字部分转换为整数,非数字部分保持不变
129
+ parts = [int(part) if part.isdigit() else part for part in parts]
130
+ return parts
131
+
132
+ def change_choices():
133
+ SoVITS_names, GPT_names = get_weights_names()
134
+ return {"choices": sorted(SoVITS_names,key=custom_sort_key), "__type__": "update"}, {"choices": sorted(GPT_names,key=custom_sort_key), "__type__": "update"}
135
+
136
+ p_label=None
137
+ p_uvr5=None
138
+ p_asr=None
139
+ p_denoise=None
140
+ p_tts_inference=None
141
+
142
+ def kill_proc_tree(pid, including_parent=True):
143
+ try:
144
+ parent = psutil.Process(pid)
145
+ except psutil.NoSuchProcess:
146
+ # Process already terminated
147
+ return
148
+
149
+ children = parent.children(recursive=True)
150
+ for child in children:
151
+ try:
152
+ os.kill(child.pid, signal.SIGTERM) # or signal.SIGKILL
153
+ except OSError:
154
+ pass
155
+ if including_parent:
156
+ try:
157
+ os.kill(parent.pid, signal.SIGTERM) # or signal.SIGKILL
158
+ except OSError:
159
+ pass
160
+
161
+ system=platform.system()
162
+ def kill_process(pid):
163
+ if(system=="Windows"):
164
+ cmd = "taskkill /t /f /pid %s" % pid
165
+ os.system(cmd)
166
+ else:
167
+ kill_proc_tree(pid)
168
+
169
+
170
+ def change_label(if_label,path_list):
171
+ global p_label
172
+ if(if_label==True and p_label==None):
173
+ path_list=my_utils.clean_path(path_list)
174
+ cmd = '"%s" tools/subfix_webui.py --load_list "%s" --webui_port %s --is_share %s'%(python_exec,path_list,webui_port_subfix,is_share)
175
+ yield i18n("打标工具WebUI已开启")
176
+ print(cmd)
177
+ p_label = Popen(cmd, shell=True)
178
+ elif(if_label==False and p_label!=None):
179
+ kill_process(p_label.pid)
180
+ p_label=None
181
+ yield i18n("打标工具WebUI已关闭")
182
+
183
+ def change_uvr5(if_uvr5):
184
+ global p_uvr5
185
+ if(if_uvr5==True and p_uvr5==None):
186
+ cmd = '"%s" tools/uvr5/webui.py "%s" %s %s %s'%(python_exec,infer_device,is_half,webui_port_uvr5,is_share)
187
+ yield i18n("UVR5已开启")
188
+ print(cmd)
189
+ p_uvr5 = Popen(cmd, shell=True)
190
+ elif(if_uvr5==False and p_uvr5!=None):
191
+ kill_process(p_uvr5.pid)
192
+ p_uvr5=None
193
+ yield i18n("UVR5已关闭")
194
+
195
+ def change_tts_inference(if_tts,bert_path,cnhubert_base_path,gpu_number,gpt_path,sovits_path):
196
+ global p_tts_inference
197
+ if(if_tts==True and p_tts_inference==None):
198
+ os.environ["gpt_path"]=gpt_path if "/" in gpt_path else "%s/%s"%(GPT_weight_root,gpt_path)
199
+ os.environ["sovits_path"]=sovits_path if "/"in sovits_path else "%s/%s"%(SoVITS_weight_root,sovits_path)
200
+ os.environ["cnhubert_base_path"]=cnhubert_base_path
201
+ os.environ["bert_path"]=bert_path
202
+ os.environ["_CUDA_VISIBLE_DEVICES"]=fix_gpu_number(gpu_number)
203
+ os.environ["is_half"]=str(is_half)
204
+ os.environ["infer_ttswebui"]=str(webui_port_infer_tts)
205
+ os.environ["is_share"]=str(is_share)
206
+ cmd = '"%s" GPT_SoVITS/inference_webui.py'%(python_exec)
207
+ yield i18n("TTS推理进程已开启")
208
+ print(cmd)
209
+ p_tts_inference = Popen(cmd, shell=True)
210
+ elif(if_tts==False and p_tts_inference!=None):
211
+ kill_process(p_tts_inference.pid)
212
+ p_tts_inference=None
213
+ yield i18n("TTS推理进程已关闭")
214
+
215
+ from tools.asr.config import asr_dict
216
+ def open_asr(asr_inp_dir, asr_opt_dir, asr_model, asr_model_size, asr_lang, asr_precision):
217
+ global p_asr
218
+ if(p_asr==None):
219
+ asr_inp_dir=my_utils.clean_path(asr_inp_dir)
220
+ asr_opt_dir=my_utils.clean_path(asr_opt_dir)
221
+ cmd = f'"{python_exec}" tools/asr/{asr_dict[asr_model]["path"]}'
222
+ cmd += f' -i "{asr_inp_dir}"'
223
+ cmd += f' -o "{asr_opt_dir}"'
224
+ cmd += f' -s {asr_model_size}'
225
+ cmd += f' -l {asr_lang}'
226
+ cmd += f" -p {asr_precision}"
227
+ output_file_name = os.path.basename(asr_inp_dir)
228
+ output_folder = asr_opt_dir or "output/asr_opt"
229
+ output_file_path = os.path.abspath(f'{output_folder}/{output_file_name}.list')
230
+ yield "ASR任务开启:%s"%cmd, {"__type__":"update","visible":False}, {"__type__":"update","visible":True}, {"__type__":"update"}
231
+ print(cmd)
232
+ p_asr = Popen(cmd, shell=True)
233
+ p_asr.wait()
234
+ p_asr=None
235
+ yield f"ASR任务完成, 查看终端进行下一步", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}, {"__type__":"update","value":output_file_path}
236
+ else:
237
+ yield "已有正在进行的ASR任务,需先终止才能开启下一次任务", {"__type__":"update","visible":False}, {"__type__":"update","visible":True}, {"__type__":"update"}
238
+ # return None
239
+
240
+ def close_asr():
241
+ global p_asr
242
+ if(p_asr!=None):
243
+ kill_process(p_asr.pid)
244
+ p_asr=None
245
+ return "已终止ASR进程", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
246
+ def open_denoise(denoise_inp_dir, denoise_opt_dir):
247
+ global p_denoise
248
+ if(p_denoise==None):
249
+ denoise_inp_dir=my_utils.clean_path(denoise_inp_dir)
250
+ denoise_opt_dir=my_utils.clean_path(denoise_opt_dir)
251
+ cmd = '"%s" tools/cmd-denoise.py -i "%s" -o "%s" -p %s'%(python_exec,denoise_inp_dir,denoise_opt_dir,"float16"if is_half==True else "float32")
252
+
253
+ yield "语音降噪任务开启:%s"%cmd, {"__type__":"update","visible":False}, {"__type__":"update","visible":True}, {"__type__":"update"}
254
+ print(cmd)
255
+ p_denoise = Popen(cmd, shell=True)
256
+ p_denoise.wait()
257
+ p_denoise=None
258
+ yield f"语音降噪任务完成, 查看终端进行下一步", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}, {"__type__":"update","value":denoise_opt_dir}
259
+ else:
260
+ yield "已有正在进行的语音降噪任务,需先终止才能开启下一次任务", {"__type__":"update","visible":False}, {"__type__":"update","visible":True}, {"__type__":"update"}
261
+ # return None
262
+
263
+ def close_denoise():
264
+ global p_denoise
265
+ if(p_denoise!=None):
266
+ kill_process(p_denoise.pid)
267
+ p_denoise=None
268
+ return "已终止语音降噪进程", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
269
+
270
+ p_train_SoVITS=None
271
+ def open1Ba(batch_size,total_epoch,exp_name,text_low_lr_rate,if_save_latest,if_save_every_weights,save_every_epoch,gpu_numbers1Ba,pretrained_s2G,pretrained_s2D):
272
+ global p_train_SoVITS
273
+ if(p_train_SoVITS==None):
274
+ with open("GPT_SoVITS/configs/s2.json")as f:
275
+ data=f.read()
276
+ data=json.loads(data)
277
+ s2_dir="%s/%s"%(exp_root,exp_name)
278
+ os.makedirs("%s/logs_s2"%(s2_dir),exist_ok=True)
279
+ if(is_half==False):
280
+ data["train"]["fp16_run"]=False
281
+ batch_size=max(1,batch_size//2)
282
+ data["train"]["batch_size"]=batch_size
283
+ data["train"]["epochs"]=total_epoch
284
+ data["train"]["text_low_lr_rate"]=text_low_lr_rate
285
+ data["train"]["pretrained_s2G"]=pretrained_s2G
286
+ data["train"]["pretrained_s2D"]=pretrained_s2D
287
+ data["train"]["if_save_latest"]=if_save_latest
288
+ data["train"]["if_save_every_weights"]=if_save_every_weights
289
+ data["train"]["save_every_epoch"]=save_every_epoch
290
+ data["train"]["gpu_numbers"]=gpu_numbers1Ba
291
+ data["data"]["exp_dir"]=data["s2_ckpt_dir"]=s2_dir
292
+ data["save_weight_dir"]=SoVITS_weight_root
293
+ data["name"]=exp_name
294
+ tmp_config_path="%s/tmp_s2.json"%tmp
295
+ with open(tmp_config_path,"w")as f:f.write(json.dumps(data))
296
+
297
+ cmd = '"%s" GPT_SoVITS/s2_train.py --config "%s"'%(python_exec,tmp_config_path)
298
+ yield "SoVITS训练开始:%s"%cmd, {"__type__":"update","visible":False}, {"__type__":"update","visible":True}
299
+ print(cmd)
300
+ p_train_SoVITS = Popen(cmd, shell=True)
301
+ p_train_SoVITS.wait()
302
+ p_train_SoVITS=None
303
+ yield "SoVITS训练完成", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
304
+ else:
305
+ yield "已有正在进行的SoVITS训练任务,需先终止才能开启下一次任务", {"__type__":"update","visible":False}, {"__type__":"update","visible":True}
306
+
307
+ def close1Ba():
308
+ global p_train_SoVITS
309
+ if(p_train_SoVITS!=None):
310
+ kill_process(p_train_SoVITS.pid)
311
+ p_train_SoVITS=None
312
+ return "已终止SoVITS训练", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
313
+
314
+ p_train_GPT=None
315
+ def open1Bb(batch_size,total_epoch,exp_name,if_dpo,if_save_latest,if_save_every_weights,save_every_epoch,gpu_numbers,pretrained_s1):
316
+ global p_train_GPT
317
+ if(p_train_GPT==None):
318
+ with open("GPT_SoVITS/configs/s1longer.yaml"if version=="v1"else "GPT_SoVITS/configs/s1longer-v2.yaml")as f:
319
+ data=f.read()
320
+ data=yaml.load(data, Loader=yaml.FullLoader)
321
+ s1_dir="%s/%s"%(exp_root,exp_name)
322
+ os.makedirs("%s/logs_s1"%(s1_dir),exist_ok=True)
323
+ if(is_half==False):
324
+ data["train"]["precision"]="32"
325
+ batch_size = max(1, batch_size // 2)
326
+ data["train"]["batch_size"]=batch_size
327
+ data["train"]["epochs"]=total_epoch
328
+ data["pretrained_s1"]=pretrained_s1
329
+ data["train"]["save_every_n_epoch"]=save_every_epoch
330
+ data["train"]["if_save_every_weights"]=if_save_every_weights
331
+ data["train"]["if_save_latest"]=if_save_latest
332
+ data["train"]["if_dpo"]=if_dpo
333
+ data["train"]["half_weights_save_dir"]=GPT_weight_root
334
+ data["train"]["exp_name"]=exp_name
335
+ data["train_semantic_path"]="%s/6-name2semantic.tsv"%s1_dir
336
+ data["train_phoneme_path"]="%s/2-name2text.txt"%s1_dir
337
+ data["output_dir"]="%s/logs_s1"%s1_dir
338
+
339
+ os.environ["_CUDA_VISIBLE_DEVICES"]=fix_gpu_numbers(gpu_numbers.replace("-",","))
340
+ os.environ["hz"]="25hz"
341
+ tmp_config_path="%s/tmp_s1.yaml"%tmp
342
+ with open(tmp_config_path, "w") as f:f.write(yaml.dump(data, default_flow_style=False))
343
+ # cmd = '"%s" GPT_SoVITS/s1_train.py --config_file "%s" --train_semantic_path "%s/6-name2semantic.tsv" --train_phoneme_path "%s/2-name2text.txt" --output_dir "%s/logs_s1"'%(python_exec,tmp_config_path,s1_dir,s1_dir,s1_dir)
344
+ cmd = '"%s" GPT_SoVITS/s1_train.py --config_file "%s" '%(python_exec,tmp_config_path)
345
+ yield "GPT训练开始:%s"%cmd, {"__type__":"update","visible":False}, {"__type__":"update","visible":True}
346
+ print(cmd)
347
+ p_train_GPT = Popen(cmd, shell=True)
348
+ p_train_GPT.wait()
349
+ p_train_GPT=None
350
+ yield "GPT训练完成", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
351
+ else:
352
+ yield "已有正在进行的GPT训练任务,需先终止才能开启下一次任务", {"__type__":"update","visible":False}, {"__type__":"update","visible":True}
353
+
354
+ def close1Bb():
355
+ global p_train_GPT
356
+ if(p_train_GPT!=None):
357
+ kill_process(p_train_GPT.pid)
358
+ p_train_GPT=None
359
+ return "已终止GPT训练", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
360
+
361
+ ps_slice=[]
362
+ def open_slice(inp,opt_root,threshold,min_length,min_interval,hop_size,max_sil_kept,_max,alpha,n_parts):
363
+ global ps_slice
364
+ inp = my_utils.clean_path(inp)
365
+ opt_root = my_utils.clean_path(opt_root)
366
+ if(os.path.exists(inp)==False):
367
+ yield "输入路径不存在", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}, {"__type__": "update"}, {"__type__": "update"}
368
+ return
369
+ if os.path.isfile(inp):n_parts=1
370
+ elif os.path.isdir(inp):pass
371
+ else:
372
+ yield "输入路径存在但既不是文件也不是文件夹", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}, {"__type__": "update"}, {"__type__": "update"}
373
+ return
374
+ if (ps_slice == []):
375
+ for i_part in range(n_parts):
376
+ cmd = '"%s" tools/slice_audio.py "%s" "%s" %s %s %s %s %s %s %s %s %s''' % (python_exec,inp, opt_root, threshold, min_length, min_interval, hop_size, max_sil_kept, _max, alpha, i_part, n_parts)
377
+ print(cmd)
378
+ p = Popen(cmd, shell=True)
379
+ ps_slice.append(p)
380
+ yield "切割执行中", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}, {"__type__": "update"}, {"__type__": "update"}
381
+ for p in ps_slice:
382
+ p.wait()
383
+ ps_slice=[]
384
+ yield "切割结束", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}, {"__type__": "update", "value":opt_root}, {"__type__": "update", "value":opt_root}
385
+ else:
386
+ yield "已有正在进行的切割任务,需先终止才能开启下一次任务", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}, {"__type__": "update"}, {"__type__": "update"}
387
+
388
+ def close_slice():
389
+ global ps_slice
390
+ if (ps_slice != []):
391
+ for p_slice in ps_slice:
392
+ try:
393
+ kill_process(p_slice.pid)
394
+ except:
395
+ traceback.print_exc()
396
+ ps_slice=[]
397
+ return "已终止所有切割进程", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
398
+
399
+ ps1a=[]
400
+ def open1a(inp_text,inp_wav_dir,exp_name,gpu_numbers,bert_pretrained_dir):
401
+ global ps1a
402
+ inp_text = my_utils.clean_path(inp_text)
403
+ inp_wav_dir = my_utils.clean_path(inp_wav_dir)
404
+ if (ps1a == []):
405
+ opt_dir="%s/%s"%(exp_root,exp_name)
406
+ config={
407
+ "inp_text":inp_text,
408
+ "inp_wav_dir":inp_wav_dir,
409
+ "exp_name":exp_name,
410
+ "opt_dir":opt_dir,
411
+ "bert_pretrained_dir":bert_pretrained_dir,
412
+ }
413
+ gpu_names=gpu_numbers.split("-")
414
+ all_parts=len(gpu_names)
415
+ for i_part in range(all_parts):
416
+ config.update(
417
+ {
418
+ "i_part": str(i_part),
419
+ "all_parts": str(all_parts),
420
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
421
+ "is_half": str(is_half)
422
+ }
423
+ )
424
+ os.environ.update(config)
425
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/1-get-text.py'%python_exec
426
+ print(cmd)
427
+ p = Popen(cmd, shell=True)
428
+ ps1a.append(p)
429
+ yield "文本进程执行中", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
430
+ for p in ps1a:
431
+ p.wait()
432
+ opt = []
433
+ for i_part in range(all_parts):
434
+ txt_path = "%s/2-name2text-%s.txt" % (opt_dir, i_part)
435
+ with open(txt_path, "r", encoding="utf8") as f:
436
+ opt += f.read().strip("\n").split("\n")
437
+ os.remove(txt_path)
438
+ path_text = "%s/2-name2text.txt" % opt_dir
439
+ with open(path_text, "w", encoding="utf8") as f:
440
+ f.write("\n".join(opt) + "\n")
441
+ ps1a=[]
442
+ if len("".join(opt)) > 0:
443
+ yield "文本进程成功", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
444
+ else:
445
+ yield "文本进程失败", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
446
+ else:
447
+ yield "已有正在进行的文本任务,需先终止才能开启下一次任务", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
448
+
449
+ def close1a():
450
+ global ps1a
451
+ if (ps1a != []):
452
+ for p1a in ps1a:
453
+ try:
454
+ kill_process(p1a.pid)
455
+ except:
456
+ traceback.print_exc()
457
+ ps1a=[]
458
+ return "已终止所有1a进程", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
459
+
460
+ ps1b=[]
461
+ def open1b(inp_text,inp_wav_dir,exp_name,gpu_numbers,ssl_pretrained_dir):
462
+ global ps1b
463
+ inp_text = my_utils.clean_path(inp_text)
464
+ inp_wav_dir = my_utils.clean_path(inp_wav_dir)
465
+ if (ps1b == []):
466
+ config={
467
+ "inp_text":inp_text,
468
+ "inp_wav_dir":inp_wav_dir,
469
+ "exp_name":exp_name,
470
+ "opt_dir":"%s/%s"%(exp_root,exp_name),
471
+ "cnhubert_base_dir":ssl_pretrained_dir,
472
+ "is_half": str(is_half)
473
+ }
474
+ gpu_names=gpu_numbers.split("-")
475
+ all_parts=len(gpu_names)
476
+ for i_part in range(all_parts):
477
+ config.update(
478
+ {
479
+ "i_part": str(i_part),
480
+ "all_parts": str(all_parts),
481
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
482
+ }
483
+ )
484
+ os.environ.update(config)
485
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/2-get-hubert-wav32k.py'%python_exec
486
+ print(cmd)
487
+ p = Popen(cmd, shell=True)
488
+ ps1b.append(p)
489
+ yield "SSL提取进程执行中", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
490
+ for p in ps1b:
491
+ p.wait()
492
+ ps1b=[]
493
+ yield "SSL提取进程结束", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
494
+ else:
495
+ yield "已有正在进行的SSL提取任务,需先终止才能开启下一次任务", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
496
+
497
+ def close1b():
498
+ global ps1b
499
+ if (ps1b != []):
500
+ for p1b in ps1b:
501
+ try:
502
+ kill_process(p1b.pid)
503
+ except:
504
+ traceback.print_exc()
505
+ ps1b=[]
506
+ return "已终止所有1b进程", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
507
+
508
+ ps1c=[]
509
+ def open1c(inp_text,exp_name,gpu_numbers,pretrained_s2G_path):
510
+ global ps1c
511
+ inp_text = my_utils.clean_path(inp_text)
512
+ if (ps1c == []):
513
+ opt_dir="%s/%s"%(exp_root,exp_name)
514
+ config={
515
+ "inp_text":inp_text,
516
+ "exp_name":exp_name,
517
+ "opt_dir":opt_dir,
518
+ "pretrained_s2G":pretrained_s2G_path,
519
+ "s2config_path":"GPT_SoVITS/configs/s2.json",
520
+ "is_half": str(is_half)
521
+ }
522
+ gpu_names=gpu_numbers.split("-")
523
+ all_parts=len(gpu_names)
524
+ for i_part in range(all_parts):
525
+ config.update(
526
+ {
527
+ "i_part": str(i_part),
528
+ "all_parts": str(all_parts),
529
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
530
+ }
531
+ )
532
+ os.environ.update(config)
533
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/3-get-semantic.py'%python_exec
534
+ print(cmd)
535
+ p = Popen(cmd, shell=True)
536
+ ps1c.append(p)
537
+ yield "语义token提取进程执行中", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
538
+ for p in ps1c:
539
+ p.wait()
540
+ opt = ["item_name\tsemantic_audio"]
541
+ path_semantic = "%s/6-name2semantic.tsv" % opt_dir
542
+ for i_part in range(all_parts):
543
+ semantic_path = "%s/6-name2semantic-%s.tsv" % (opt_dir, i_part)
544
+ with open(semantic_path, "r", encoding="utf8") as f:
545
+ opt += f.read().strip("\n").split("\n")
546
+ os.remove(semantic_path)
547
+ with open(path_semantic, "w", encoding="utf8") as f:
548
+ f.write("\n".join(opt) + "\n")
549
+ ps1c=[]
550
+ yield "语义token提取进程结束", {"__type__":"update","visible":True}, {"__type__":"update","visible":False}
551
+ else:
552
+ yield "已有正在进行的语义token提取任务,需先终止才能开启下一次任务", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
553
+
554
+ def close1c():
555
+ global ps1c
556
+ if (ps1c != []):
557
+ for p1c in ps1c:
558
+ try:
559
+ kill_process(p1c.pid)
560
+ except:
561
+ traceback.print_exc()
562
+ ps1c=[]
563
+ return "已终止所有语义token进程", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
564
+ #####inp_text,inp_wav_dir,exp_name,gpu_numbers1a,gpu_numbers1Ba,gpu_numbers1c,bert_pretrained_dir,cnhubert_base_dir,pretrained_s2G
565
+ ps1abc=[]
566
+ def open1abc(inp_text,inp_wav_dir,exp_name,gpu_numbers1a,gpu_numbers1Ba,gpu_numbers1c,bert_pretrained_dir,ssl_pretrained_dir,pretrained_s2G_path):
567
+ global ps1abc
568
+ inp_text = my_utils.clean_path(inp_text)
569
+ inp_wav_dir = my_utils.clean_path(inp_wav_dir)
570
+ if (ps1abc == []):
571
+ opt_dir="%s/%s"%(exp_root,exp_name)
572
+ try:
573
+ #############################1a
574
+ path_text="%s/2-name2text.txt" % opt_dir
575
+ if(os.path.exists(path_text)==False or (os.path.exists(path_text)==True and len(open(path_text,"r",encoding="utf8").read().strip("\n").split("\n"))<2)):
576
+ config={
577
+ "inp_text":inp_text,
578
+ "inp_wav_dir":inp_wav_dir,
579
+ "exp_name":exp_name,
580
+ "opt_dir":opt_dir,
581
+ "bert_pretrained_dir":bert_pretrained_dir,
582
+ "is_half": str(is_half)
583
+ }
584
+ gpu_names=gpu_numbers1a.split("-")
585
+ all_parts=len(gpu_names)
586
+ for i_part in range(all_parts):
587
+ config.update(
588
+ {
589
+ "i_part": str(i_part),
590
+ "all_parts": str(all_parts),
591
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
592
+ }
593
+ )
594
+ os.environ.update(config)
595
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/1-get-text.py'%python_exec
596
+ print(cmd)
597
+ p = Popen(cmd, shell=True)
598
+ ps1abc.append(p)
599
+ yield "进度:1a-ing", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
600
+ for p in ps1abc:p.wait()
601
+
602
+ opt = []
603
+ for i_part in range(all_parts):#txt_path="%s/2-name2text-%s.txt"%(opt_dir,i_part)
604
+ txt_path = "%s/2-name2text-%s.txt" % (opt_dir, i_part)
605
+ with open(txt_path, "r",encoding="utf8") as f:
606
+ opt += f.read().strip("\n").split("\n")
607
+ os.remove(txt_path)
608
+ with open(path_text, "w",encoding="utf8") as f:
609
+ f.write("\n".join(opt) + "\n")
610
+ assert len("".join(opt)) > 0, "1Aa-文本获取进程失败"
611
+ yield "进度:1a-done", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
612
+ ps1abc=[]
613
+ #############################1b
614
+ config={
615
+ "inp_text":inp_text,
616
+ "inp_wav_dir":inp_wav_dir,
617
+ "exp_name":exp_name,
618
+ "opt_dir":opt_dir,
619
+ "cnhubert_base_dir":ssl_pretrained_dir,
620
+ }
621
+ gpu_names=gpu_numbers1Ba.split("-")
622
+ all_parts=len(gpu_names)
623
+ for i_part in range(all_parts):
624
+ config.update(
625
+ {
626
+ "i_part": str(i_part),
627
+ "all_parts": str(all_parts),
628
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
629
+ }
630
+ )
631
+ os.environ.update(config)
632
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/2-get-hubert-wav32k.py'%python_exec
633
+ print(cmd)
634
+ p = Popen(cmd, shell=True)
635
+ ps1abc.append(p)
636
+ yield "进度:1a-done, 1b-ing", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
637
+ for p in ps1abc:p.wait()
638
+ yield "进度:1a1b-done", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
639
+ ps1abc=[]
640
+ #############################1c
641
+ path_semantic = "%s/6-name2semantic.tsv" % opt_dir
642
+ if(os.path.exists(path_semantic)==False or (os.path.exists(path_semantic)==True and os.path.getsize(path_semantic)<31)):
643
+ config={
644
+ "inp_text":inp_text,
645
+ "exp_name":exp_name,
646
+ "opt_dir":opt_dir,
647
+ "pretrained_s2G":pretrained_s2G_path,
648
+ "s2config_path":"GPT_SoVITS/configs/s2.json",
649
+ }
650
+ gpu_names=gpu_numbers1c.split("-")
651
+ all_parts=len(gpu_names)
652
+ for i_part in range(all_parts):
653
+ config.update(
654
+ {
655
+ "i_part": str(i_part),
656
+ "all_parts": str(all_parts),
657
+ "_CUDA_VISIBLE_DEVICES": fix_gpu_number(gpu_names[i_part]),
658
+ }
659
+ )
660
+ os.environ.update(config)
661
+ cmd = '"%s" GPT_SoVITS/prepare_datasets/3-get-semantic.py'%python_exec
662
+ print(cmd)
663
+ p = Popen(cmd, shell=True)
664
+ ps1abc.append(p)
665
+ yield "进度:1a1b-done, 1cing", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
666
+ for p in ps1abc:p.wait()
667
+
668
+ opt = ["item_name\tsemantic_audio"]
669
+ for i_part in range(all_parts):
670
+ semantic_path = "%s/6-name2semantic-%s.tsv" % (opt_dir, i_part)
671
+ with open(semantic_path, "r",encoding="utf8") as f:
672
+ opt += f.read().strip("\n").split("\n")
673
+ os.remove(semantic_path)
674
+ with open(path_semantic, "w",encoding="utf8") as f:
675
+ f.write("\n".join(opt) + "\n")
676
+ yield "进度:all-done", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
677
+ ps1abc = []
678
+ yield "一键三连进程结束", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
679
+ except:
680
+ traceback.print_exc()
681
+ close1abc()
682
+ yield "一键三连中途报错", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
683
+ else:
684
+ yield "已有正在进行的一键三连任务,需先终止才能开启下一次任务", {"__type__": "update", "visible": False}, {"__type__": "update", "visible": True}
685
+
686
+ def close1abc():
687
+ global ps1abc
688
+ if (ps1abc != []):
689
+ for p1abc in ps1abc:
690
+ try:
691
+ kill_process(p1abc.pid)
692
+ except:
693
+ traceback.print_exc()
694
+ ps1abc=[]
695
+ return "已终止所有一键三连进程", {"__type__": "update", "visible": True}, {"__type__": "update", "visible": False}
696
+
697
+ with gr.Blocks(title="GPT-SoVITS WebUI") as app:
698
+ gr.Markdown(
699
+ value=
700
+ i18n("本软件以MIT协议开源, 作者不对软件具备任何控制力, 使用软件者、传播软件导出的声音者自负全责. <br>如不认可该条款, 则不能使用或引用软件包内任何代码和文件. 详见根目录<b>LICENSE</b>.")
701
+ )
702
+ gr.Markdown(
703
+ value=
704
+ i18n("中文教程文档:https://www.yuque.com/baicaigongchang1145haoyuangong/ib3g1e")
705
+ )
706
+
707
+ with gr.Tabs():
708
+ with gr.TabItem(i18n("0-前置数据集获取工具")):#提前随机切片防止uvr5爆内存->uvr5->slicer->asr->打标
709
+ gr.Markdown(value=i18n("0a-UVR5人声伴奏分离&去混响去延迟工具"))
710
+ with gr.Row():
711
+ if_uvr5 = gr.Checkbox(label=i18n("是否开启UVR5-WebUI"),show_label=True)
712
+ uvr5_info = gr.Textbox(label=i18n("UVR5进程输出信息"))
713
+ gr.Markdown(value=i18n("0b-语音切分工具"))
714
+ with gr.Row():
715
+ with gr.Column(scale=3):
716
+ with gr.Row():
717
+ slice_inp_path=gr.Textbox(label=i18n("音频自动切分输入路径,可文件可文件夹"),value="")
718
+ slice_opt_root=gr.Textbox(label=i18n("切分后的子音频的输出根目录"),value="output/slicer_opt")
719
+ with gr.Row():
720
+ threshold=gr.Textbox(label=i18n("threshold:音量小于这个值视作静音的备选切割点"),value="-34")
721
+ min_length=gr.Textbox(label=i18n("min_length:每段最小多长,如果第一段太短一直和后面段连起来直到超过这个值"),value="4000")
722
+ min_interval=gr.Textbox(label=i18n("min_interval:最短切割间隔"),value="300")
723
+ hop_size=gr.Textbox(label=i18n("hop_size:怎么算音量曲线,越小精度越大计算量越高(不是精度越大效果越好)"),value="10")
724
+ max_sil_kept=gr.Textbox(label=i18n("max_sil_kept:切完后静音最多留多长"),value="500")
725
+ with gr.Row():
726
+ _max=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("max:归一化后最大值多少"),value=0.9,interactive=True)
727
+ alpha=gr.Slider(minimum=0,maximum=1,step=0.05,label=i18n("alpha_mix:混多少比例归一化后音频进来"),value=0.25,interactive=True)
728
+ n_process=gr.Slider(minimum=1,maximum=n_cpu,step=1,label=i18n("切割使用的进程数"),value=4,interactive=True)
729
+ with gr.Row():
730
+ slicer_info = gr.Textbox(label=i18n("语音切割进程输出信息"))
731
+ open_slicer_button=gr.Button(i18n("开启语音切割"), variant="primary",visible=True)
732
+ close_slicer_button=gr.Button(i18n("终止语音切割"), variant="primary",visible=False)
733
+ gr.Markdown(value=i18n("0bb-语音降噪工具"))
734
+ with gr.Row():
735
+ with gr.Column(scale=3):
736
+ with gr.Row():
737
+ denoise_input_dir=gr.Textbox(label=i18n("降噪音频文件输入文件夹"),value="")
738
+ denoise_output_dir=gr.Textbox(label=i18n("降噪结果输出文件夹"),value="output/denoise_opt")
739
+ with gr.Row():
740
+ denoise_info = gr.Textbox(label=i18n("语音降噪进程输出信息"))
741
+ open_denoise_button = gr.Button(i18n("开启语音降噪"), variant="primary",visible=True)
742
+ close_denoise_button = gr.Button(i18n("终止语音降噪进程"), variant="primary",visible=False)
743
+ gr.Markdown(value=i18n("0c-中文批量离线ASR工具"))
744
+ with gr.Row():
745
+ with gr.Column(scale=3):
746
+ with gr.Row():
747
+ asr_inp_dir = gr.Textbox(
748
+ label=i18n("输入文件夹路径"),
749
+ value="D:\\GPT-SoVITS\\raw\\xxx",
750
+ interactive=True,
751
+ )
752
+ asr_opt_dir = gr.Textbox(
753
+ label = i18n("输出文件夹路径"),
754
+ value = "output/asr_opt",
755
+ interactive = True,
756
+ )
757
+ with gr.Row():
758
+ asr_model = gr.Dropdown(
759
+ label = i18n("ASR 模型"),
760
+ choices = list(asr_dict.keys()),
761
+ interactive = True,
762
+ value="达摩 ASR (中文)"
763
+ )
764
+ asr_size = gr.Dropdown(
765
+ label = i18n("ASR 模型尺寸"),
766
+ choices = ["large"],
767
+ interactive = True,
768
+ value="large"
769
+ )
770
+ asr_lang = gr.Dropdown(
771
+ label = i18n("ASR 语言设置"),
772
+ choices = ["zh"],
773
+ interactive = True,
774
+ value="zh"
775
+ )
776
+ asr_precision = gr.Dropdown(
777
+ label = i18n("数据类型精度"),
778
+ choices = ["float32"],
779
+ interactive = True,
780
+ value="float32"
781
+ )
782
+ with gr.Row():
783
+ asr_info = gr.Textbox(label=i18n("ASR进程输出信息"))
784
+ open_asr_button = gr.Button(i18n("开启离线批量ASR"), variant="primary",visible=True)
785
+ close_asr_button = gr.Button(i18n("终止ASR进程"), variant="primary",visible=False)
786
+
787
+ def change_lang_choices(key): #根据选择的模型修改可选的语言
788
+ # return gr.Dropdown(choices=asr_dict[key]['lang'])
789
+ return {"__type__": "update", "choices": asr_dict[key]['lang'],"value":asr_dict[key]['lang'][0]}
790
+ def change_size_choices(key): # 根据选择的模型修改可选的模型尺寸
791
+ # return gr.Dropdown(choices=asr_dict[key]['size'])
792
+ return {"__type__": "update", "choices": asr_dict[key]['size'],"value":asr_dict[key]['size'][-1]}
793
+ def change_precision_choices(key): #根据选择的模型修改可选的语言
794
+ if key =="Faster Whisper (多语种)":
795
+ if default_batch_size <= 4:
796
+ precision = 'int8'
797
+ elif is_half:
798
+ precision = 'float16'
799
+ else:
800
+ precision = 'float32'
801
+ else:
802
+ precision = 'float32'
803
+ # return gr.Dropdown(choices=asr_dict[key]['precision'])
804
+ return {"__type__": "update", "choices": asr_dict[key]['precision'],"value":precision}
805
+ asr_model.change(change_lang_choices, [asr_model], [asr_lang])
806
+ asr_model.change(change_size_choices, [asr_model], [asr_size])
807
+ asr_model.change(change_precision_choices, [asr_model], [asr_precision])
808
+
809
+
810
+ gr.Markdown(value=i18n("0d-语音文本校对标注工具"))
811
+ with gr.Row():
812
+ if_label = gr.Checkbox(label=i18n("是否开启打标WebUI"),show_label=True)
813
+ path_list = gr.Textbox(
814
+ label=i18n(".list标注文件的路径"),
815
+ value="D:\\RVC1006\\GPT-SoVITS\\raw\\xxx.list",
816
+ interactive=True,
817
+ )
818
+ label_info = gr.Textbox(label=i18n("打标工具进程输出信息"))
819
+ if_label.change(change_label, [if_label,path_list], [label_info])
820
+ if_uvr5.change(change_uvr5, [if_uvr5], [uvr5_info])
821
+ open_asr_button.click(open_asr, [asr_inp_dir, asr_opt_dir, asr_model, asr_size, asr_lang, asr_precision], [asr_info,open_asr_button,close_asr_button,path_list])
822
+ close_asr_button.click(close_asr, [], [asr_info,open_asr_button,close_asr_button])
823
+ open_slicer_button.click(open_slice, [slice_inp_path,slice_opt_root,threshold,min_length,min_interval,hop_size,max_sil_kept,_max,alpha,n_process], [slicer_info,open_slicer_button,close_slicer_button,asr_inp_dir,denoise_input_dir])
824
+ close_slicer_button.click(close_slice, [], [slicer_info,open_slicer_button,close_slicer_button])
825
+ open_denoise_button.click(open_denoise, [denoise_input_dir,denoise_output_dir], [denoise_info,open_denoise_button,close_denoise_button,asr_inp_dir])
826
+ close_denoise_button.click(close_denoise, [], [denoise_info,open_denoise_button,close_denoise_button])
827
+
828
+ with gr.TabItem(i18n("1-GPT-SoVITS-TTS")):
829
+ with gr.Row():
830
+ exp_name = gr.Textbox(label=i18n("*实验/模型名"), value="xxx", interactive=True)
831
+ gpu_info = gr.Textbox(label=i18n("显卡信息"), value=gpu_info, visible=True, interactive=False)
832
+ pretrained_s2G = gr.Textbox(label=i18n("预训练的SoVITS-G模型路径"), value=pretrained_sovits_name, interactive=True)
833
+ pretrained_s2D = gr.Textbox(label=i18n("预训练的SoVITS-D模型路径"), value=pretrained_sovits_name.replace("s2G","s2D"), interactive=True)
834
+ pretrained_s1 = gr.Textbox(label=i18n("预训练的GPT模型路径"), value=pretrained_gpt_name, interactive=True)
835
+ with gr.TabItem(i18n("1A-训练集格式化工具")):
836
+ gr.Markdown(value=i18n("输出logs/实验名目录下应有23456开头的文件和文件夹"))
837
+ with gr.Row():
838
+ inp_text = gr.Textbox(label=i18n("*文本标注文件"),value=r"D:\RVC1006\GPT-SoVITS\raw\xxx.list",interactive=True)
839
+ inp_wav_dir = gr.Textbox(
840
+ label=i18n("*训练集音频文件目录"),
841
+ # value=r"D:\RVC1006\GPT-SoVITS\raw\xxx",
842
+ interactive=True,
843
+ placeholder=i18n("填切割后音频所在目录!读取的音频文件完整路径=该目录-拼接-list文件里波形对应的文件名(不是全路径)。如果留空则使用.list文件里的绝对全路径。")
844
+ )
845
+ gr.Markdown(value=i18n("1Aa-文本内容"))
846
+ with gr.Row():
847
+ gpu_numbers1a = gr.Textbox(label=i18n("GPU卡号以-分割,每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
848
+ bert_pretrained_dir = gr.Textbox(label=i18n("预训练的中文BERT模型路径"),value="GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",interactive=False)
849
+ button1a_open = gr.Button(i18n("开启文本获取"), variant="primary",visible=True)
850
+ button1a_close = gr.Button(i18n("终止文本获取进程"), variant="primary",visible=False)
851
+ info1a=gr.Textbox(label=i18n("文本进程输出信息"))
852
+ gr.Markdown(value=i18n("1Ab-SSL自监督特征提取"))
853
+ with gr.Row():
854
+ gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割,每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
855
+ cnhubert_base_dir = gr.Textbox(label=i18n("预训练的SSL模型路径"),value="GPT_SoVITS/pretrained_models/chinese-hubert-base",interactive=False)
856
+ button1b_open = gr.Button(i18n("开启SSL提取"), variant="primary",visible=True)
857
+ button1b_close = gr.Button(i18n("终止SSL提取进程"), variant="primary",visible=False)
858
+ info1b=gr.Textbox(label=i18n("SSL进程输出信息"))
859
+ gr.Markdown(value=i18n("1Ac-语义token提取"))
860
+ with gr.Row():
861
+ gpu_numbers1c = gr.Textbox(label=i18n("GPU卡号以-分割,每个卡号一个进程"),value="%s-%s"%(gpus,gpus),interactive=True)
862
+ button1c_open = gr.Button(i18n("开启语义token提取"), variant="primary",visible=True)
863
+ button1c_close = gr.Button(i18n("终止语义token提取进程"), variant="primary",visible=False)
864
+ info1c=gr.Textbox(label=i18n("语义token提取进程输出信息"))
865
+ gr.Markdown(value=i18n("1Aabc-训练集格式化一键三连"))
866
+ with gr.Row():
867
+ button1abc_open = gr.Button(i18n("开启一键三连"), variant="primary",visible=True)
868
+ button1abc_close = gr.Button(i18n("终止一键三连"), variant="primary",visible=False)
869
+ info1abc=gr.Textbox(label=i18n("一键三连进程输出信息"))
870
+ button1a_open.click(open1a, [inp_text,inp_wav_dir,exp_name,gpu_numbers1a,bert_pretrained_dir], [info1a,button1a_open,button1a_close])
871
+ button1a_close.click(close1a, [], [info1a,button1a_open,button1a_close])
872
+ button1b_open.click(open1b, [inp_text,inp_wav_dir,exp_name,gpu_numbers1Ba,cnhubert_base_dir], [info1b,button1b_open,button1b_close])
873
+ button1b_close.click(close1b, [], [info1b,button1b_open,button1b_close])
874
+ button1c_open.click(open1c, [inp_text,exp_name,gpu_numbers1c,pretrained_s2G], [info1c,button1c_open,button1c_close])
875
+ button1c_close.click(close1c, [], [info1c,button1c_open,button1c_close])
876
+ button1abc_open.click(open1abc, [inp_text,inp_wav_dir,exp_name,gpu_numbers1a,gpu_numbers1Ba,gpu_numbers1c,bert_pretrained_dir,cnhubert_base_dir,pretrained_s2G], [info1abc,button1abc_open,button1abc_close])
877
+ button1abc_close.click(close1abc, [], [info1abc,button1abc_open,button1abc_close])
878
+ with gr.TabItem(i18n("1B-微调训练")):
879
+ gr.Markdown(value=i18n("1Ba-SoVITS训练。用于分享的模型文件输出在SoVITS_weights下。"))
880
+ with gr.Row():
881
+ batch_size = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
882
+ total_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("总训练轮数total_epoch,不建议太高"),value=8,interactive=True)
883
+ text_low_lr_rate = gr.Slider(minimum=0.2,maximum=0.6,step=0.05,label=i18n("文本模块学习率权重"),value=0.4,interactive=True)
884
+ save_every_epoch = gr.Slider(minimum=1,maximum=25,step=1,label=i18n("保存频率save_every_epoch"),value=4,interactive=True)
885
+ if_save_latest = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
886
+ if_save_every_weights = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
887
+ gpu_numbers1Ba = gr.Textbox(label=i18n("GPU卡号以-分割,每个卡号一个进程"), value="%s" % (gpus), interactive=True)
888
+ with gr.Row():
889
+ button1Ba_open = gr.Button(i18n("开启SoVITS训练"), variant="primary",visible=True)
890
+ button1Ba_close = gr.Button(i18n("终止SoVITS训练"), variant="primary",visible=False)
891
+ info1Ba=gr.Textbox(label=i18n("SoVITS训练进程输出信息"))
892
+ gr.Markdown(value=i18n("1Bb-GPT训练。用于分享的模型文件输出在GPT_weights下。"))
893
+ with gr.Row():
894
+ batch_size1Bb = gr.Slider(minimum=1,maximum=40,step=1,label=i18n("每张显卡的batch_size"),value=default_batch_size,interactive=True)
895
+ total_epoch1Bb = gr.Slider(minimum=2,maximum=50,step=1,label=i18n("总训练轮数total_epoch"),value=15,interactive=True)
896
+ if_dpo = gr.Checkbox(label=i18n("是否开启dpo训练选项(实验性)"), value=False, interactive=True, show_label=True)
897
+ if_save_latest1Bb = gr.Checkbox(label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), value=True, interactive=True, show_label=True)
898
+ if_save_every_weights1Bb = gr.Checkbox(label=i18n("是否在每次保存时间点将最终小模型保存至weights文件夹"), value=True, interactive=True, show_label=True)
899
+ save_every_epoch1Bb = gr.Slider(minimum=1,maximum=50,step=1,label=i18n("保存频率save_every_epoch"),value=5,interactive=True)
900
+ gpu_numbers1Bb = gr.Textbox(label=i18n("GPU卡号以-分割,每个卡号一个进程"), value="%s" % (gpus), interactive=True)
901
+ with gr.Row():
902
+ button1Bb_open = gr.Button(i18n("开启GPT训练"), variant="primary",visible=True)
903
+ button1Bb_close = gr.Button(i18n("终止GPT训练"), variant="primary",visible=False)
904
+ info1Bb=gr.Textbox(label=i18n("GPT训练进程输出信息"))
905
+ button1Ba_open.click(open1Ba, [batch_size,total_epoch,exp_name,text_low_lr_rate,if_save_latest,if_save_every_weights,save_every_epoch,gpu_numbers1Ba,pretrained_s2G,pretrained_s2D], [info1Ba,button1Ba_open,button1Ba_close])
906
+ button1Ba_close.click(close1Ba, [], [info1Ba,button1Ba_open,button1Ba_close])
907
+ button1Bb_open.click(open1Bb, [batch_size1Bb,total_epoch1Bb,exp_name,if_dpo,if_save_latest1Bb,if_save_every_weights1Bb,save_every_epoch1Bb,gpu_numbers1Bb,pretrained_s1], [info1Bb,button1Bb_open,button1Bb_close])
908
+ button1Bb_close.click(close1Bb, [], [info1Bb,button1Bb_open,button1Bb_close])
909
+ with gr.TabItem(i18n("1C-推理")):
910
+ gr.Markdown(value=i18n("选择训练完存放在SoVITS_weights和GPT_weights下的模型。默认的一个是底模,体验5秒Zero Shot TTS用。"))
911
+ with gr.Row():
912
+ GPT_dropdown = gr.Dropdown(label=i18n("*GPT模型列表"), choices=sorted(GPT_names,key=custom_sort_key),value=pretrained_gpt_name,interactive=True)
913
+ SoVITS_dropdown = gr.Dropdown(label=i18n("*SoVITS模型列表"), choices=sorted(SoVITS_names,key=custom_sort_key),value=pretrained_sovits_name,interactive=True)
914
+ gpu_number_1C=gr.Textbox(label=i18n("GPU卡号,只能填1个整数"), value=gpus, interactive=True)
915
+ refresh_button = gr.Button(i18n("刷新模型路径"), variant="primary")
916
+ refresh_button.click(fn=change_choices,inputs=[],outputs=[SoVITS_dropdown,GPT_dropdown])
917
+ with gr.Row():
918
+ if_tts = gr.Checkbox(label=i18n("是否开启TTS推理WebUI"), show_label=True)
919
+ tts_info = gr.Textbox(label=i18n("TTS推理WebUI进程输出信息"))
920
+ if_tts.change(change_tts_inference, [if_tts,bert_pretrained_dir,cnhubert_base_dir,gpu_number_1C,GPT_dropdown,SoVITS_dropdown], [tts_info])
921
+ with gr.TabItem(i18n("2-GPT-SoVITS-变声")):gr.Markdown(value=i18n("施工中,请静候佳音"))
922
+ # app.queue(concurrency_count=511, max_size=1022).launch(
923
+ # server_name="0.0.0.0",
924
+ # inbrowser=True,
925
+ # share=is_share,
926
+ # server_port=webui_port_main,
927
+ # quiet=True,
928
+ # )
929
+ # button1abc_open.click(open1abc, [inp_text,inp_wav_dir,exp_name,gpu_numbers1a,gpu_numbers1Ba,gpu_numbers1c,bert_pretrained_dir,cnhubert_base_dir,pretrained_s2G], [info1abc,button1abc_open,button1abc_close])
930
+ # open1abc(./*/transcript.txt, None, *, "0-1", "0-1", "0-1", "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large", "GPT_SoVITS/pretrained_models/chinese-hubert-base", "GPT_SoVITS/pretrained_models/s2G2333k.pth")
931
+
932
  import os
933
  import sys
934
  # to avoid the modified user.pth file